Potent LpxC Inhibitors with In Vitro Activity against Multidrug-Resistant Pseudomonas aeruginosa

New drugs with novel mechanisms of resistance are desperately needed to address both community and nosocomial infections due to Gram-negative bacteria. One such potential target is LpxC, an essential enzyme that catalyzes the first committed step of lipid A biosynthesis. Achaogen conducted an extensive research campaign to discover novel LpxC inhibitors with activity against Pseudomonas aeruginosa.

ABSTRACT

New drugs with novel mechanisms of resistance are desperately needed to address both community and nosocomial infections due to Gram-negative bacteria. One such potential target is LpxC, an essential enzyme that catalyzes the first committed step of lipid A biosynthesis. Achaogen conducted an extensive research campaign to discover novel LpxC inhibitors with activity against Pseudomonas aeruginosa. We report here the in vitro antibacterial activity and pharmacodynamics of ACHN-975, the only molecule from these efforts and the first ever LpxC inhibitor to be evaluated in phase 1 clinical trials. In addition, we describe the profiles of three additional LpxC inhibitors that were identified as potential lead molecules. These efforts did not produce an additional development candidate with a sufficiently large therapeutic window and the program was subsequently terminated.

INTRODUCTION

Pseudomonas aeruginosa is a common cause of health care-associated infections, including hospital-acquired pneumonia, urinary tract infections, and infected burns and wounds (https://www.cdc.gov/hai/organisms/pseudomonas.html). It is also the most common cause of chronic infection in patients with cystic fibrosis (CF) (1). Many isolates of P. aeruginosa are multidrug resistant (MDR), resulting in difficult-to-treat infections (2). P. aeruginosa organisms often carry beta-lactamases, rendering many isolates resistant to penicillins, cephalosporins, and, in some cases, carbapenems (3). In addition, overexpression of efflux pumps may play a role in the development of resistance to fluoroquinolones and aminoglycosides, whereas the ability to downregulate expression of the porin OprD can lead to carbapenem resistance (4–6). P. aeruginosa also adeptly acquires resistance genes through mutation or horizontal gene transfer, including fluoroquinolone resistance genes and genes for extended-spectrum beta-lactamases (ESBLs), carbapenemases, and aminoglycoside-modifying enzymes (3). Finally, isolates resistant to colistin, a drug of last resort, are beginning to emerge (7).

New antibiotics and novel targets will need to be discovered to address the emerging antibacterial resistance problem in P. aeruginosa. One such target is LpxC {UDP-[3-O-(R)-3-hydroxymyristoyl]-N-acetylglucosamine deacetylase}. LpxC is a zinc metalloenzyme that catalyzes the first committed step in the biosynthesis of the lipid A subunit of lipopolysaccharide (LPS), a critical component of the outer membrane of most Gram-negative bacteria, including P. aeruginosa (8, 9). It is an attractive target for new antibacterial discovery because it is essential for bacterial survival in most Gram-negative species, is highly conserved, and has no known human homologue (8, 10).

Several LpxC inhibitor discovery programs have previously been conducted, and the published results describe efforts to identify an LpxC inhibitor with drug-like properties and a sufficiently large therapeutic window to study in human clinical trials (11, 12). Our LpxC inhibitor program yielded several potential lead molecules, but only ACHN-975 reached human clinical testing. A double-blind, randomized, placebo-controlled, single-ascending-dose study was conducted to assess the safety, tolerability, and pharmacokinetics (PK) of ACHN-975 in healthy volunteers (https://clinicaltrials.gov/ct2/show/NCT01597947). This trial revealed a peak plasma concentration (C_max)-driven dose-limiting toxicity (DLT) of transient hypotension without tachycardia. The therapeutic window of ACHN-975 was insufficient for the drug to advance in development. Subsequently, we conducted a second LpxC inhibitor discovery campaign, which was focused on maintaining activity against P. aeruginosa specifically while reducing toxicity (13). Additional LpxC inhibitors were synthesized and assessed in a series of go/no-go assays, which included in vitro MIC assays, plasma protein binding, cytotoxicity studies, and a high-content cardiovascular safety assay in rats. Three new lead LpxC inhibitor molecules were identified, LPXC-289, LPXC-313, and LPXC-516, which appeared to have a wider therapeutic window than ACHN-975, and additional studies were performed; LPXC-516 was notably the most promising of the three leads due to favorable activity, safety, and pharmacokinetics (13). A solubility-enhancing prodrug of LPXC-516 was developed to enable dose escalation. However, further development of all three of these lead molecules was also halted due to reemergence of the cardiovascular toxicity in preclinical models.

It is also noteworthy that these molecules had activity against other Gram-negative bacteria, including Enterobacteriaceae and Gram-negative biothreat agents (14–16). However, unpublished in vivo investigations using isolates of P. aeruginosa, Escherichia coli, and Klebsiella pneumoniae isolates revealed apparent differences in the relationships between the PK and pharmacodynamics (PD) of the molecules against these different bacterial species. For P. aeruginosa, a C_max-driven effect, achieved through less frequent doses, appeared to provide maximal efficacy, whereas the maximal PD response against Enterobacteriaceae appeared to be driven by a time-dependent effect. There is more work to be done to completely understand this phenomenon, but the challenges in designing a dose and optimizing potency against this diverse group of pathogens led us to focus the program solely on P. aeruginosa.

RESULTS

In vitro activities of ACHN-975, LPXC-516, LPXC-313, and LPXC-289.

Figure 1 shows the structures, molecular weights, and 50% inhibitory concentration (IC₅₀) values measured against the purified P. aeruginosa LpxC enzyme for ACHN-975 and the three additional leads. Notably, all of the new compounds retained potent inhibition against the LpxC enzyme. ACHN-975 was potent against the P. aeruginosa isolates tested, inhibiting 100% of the isolates at an MIC of ≤2 μg/ml and with an MIC₅₀ and MIC₉₀ of 0.06 and 0.25 μg/ml, respectively (Table 1). LPXC-516, LPXC-313, and LPXC-289 were less potent than ACHN-975, each with MIC₉₀ values of 2 μg/ml. Overall, the P. aeruginosa isolates tested had various susceptibilities to other antimicrobials.

FIG 1.

Structures and characteristics of LpxC inhibitors. ^a, mean MIC value against 5 P. aeruginosa isolates used as part of a primary screening panel that were used to predict MIC values against larger data sets. ^b, plasma clearance measured in rats (from Cohen et al. [13]).

TABLE 1.

In vitro activities of LpxC inhibitors against Pseudomonas aeruginosa isolates

Organism(s) (n)	Antimicrobial	% Sa	% Ib	% Rc	MIC (μg/ml)
					50%	90%	Range
All P. aeruginosa isolates (250)	ACHN-975	NAf	NA	NA	0.06	0.25	≤0.008 to 2
	LPXC-516	NA	NA	NA	1	2	≤0.03 to 16
	LPXC-289	NA	NA	NA	0.25	2	≤0.015 to 8
	LPXC-313	NA	NA	NA	0.5	2	≤0.03 to 8
	Amikacin	85.2	6.4	8.4	4	32	≤0.5 to >32
	Ceftazidime	69.2	8.8	22.0	4	>16	≤0.5 to >16
	Colistin	100	0	0	1	1	≤0.25 to 2
	Levofloxacin	61.6	10.8	27.6	1	>8	≤0.25 to >8
	Meropenem	72.8	8.1	18.8	0.5	>8	≤0.12 to >8
	PIP-TAZd	66.4	13.6	20.0	8	>128	≤2 to >128
	Tobramycin	82.0	3.6	14.4	0.5	>8	≤0.12 to >8
CF P. aeruginosa isolates (49)e	ACHN-975	NA	NA	NA	0.06	0.25	0.015 to 1
	LPXC-516	NA	NA	NA	0.5	2	0.12 to 2
	LPXC-289	NA	NA	NA	0.12	1	≤0.015 to 2
	LPXC-313	NA	NA	NA	0.25	2	≤0.03 to 2
	Amikacin	61.2	18.4	20.4	16	>32	2 to >32
	Ceftazidime	51.0	16.3	32.7	8	>16	1 to >16
	Colistin	100	0	0	1	1	≤0.25 to 1
	Levofloxacin	40.8	24.5	34.7	4	>8	≤0.25 to >8
	Meropenem	83.7	6.1	10.2	0.5	8	≤0.12 to >8
	PIP-TAZd	49.0	24.5	26.5	32	>128	≤2 to >128
	Tobramycin	73.5	10.2	16.3	2	>8	0.25 to >8
Non-CF P. aeruginosa isolates (201)	ACHN-975	NA	NA	NA	0.06	0.25	≤0.008 to 2
	LPXC-516	NA	NA	NA	1	2	≤0.03 to 16
	LPXC-289	NA	NA	NA	0.25	2	≤0.015 to 8
	LPXC-313	NA	NA	NA	0.5	2	≤0.03 to 8
	Amikacin	91.0	3.5	5.5	4	16	≤0.5 to >32
	Ceftazidime	73.6	7.0	19.4	2	>16	≤0.5 to >16
	Colistin	100	0	0	1	1	≤0.25 to 2
	Levofloxacin	66.7	7.5	25.9	1	>8	≤0.25 to >8
	Meropenem	70.2	9.0	20.9	0.5	>8	≤0.12 to >8
	PIP-TAZd	70.7	11.0	18.4	8	>128	≤2 to >128
	Tobramycin	84.1	2.0	13.9	0.5	>8	≤0.12 to >8

^aPercentage of isolates susceptible (S) per CLSI interpretive criteria.

^bPercentage of isolates intermediate (I) per CLSI interpretive criteria.

^cPercentage of isolates resistant (R) per CLSI interpretive criteria.

^dPIP-TAZ, piperacillin-tazobactam.

^eIsolates from patients with cystic fibrosis.

^fNA, not applicable.

This collection of P. aeruginosa isolates also included a subset of respiratory isolates from cystic fibrosis (CF) patients. The activity of each of the four LpxC inhibitors against the CF isolate subset was similar to the activity against the non-CF isolate subset. Specifically, the MIC₉₀ values of ACHN-975, LPXC-516, and LPXC-313 were identical between the two subsets (0.25, 2, and 2 μg/ml, respectively), while for LPXC-289, MIC₉₀ values were also similar, at 1 μg/ml for the CF isolate subset, versus 2 μg/ml for the non-CF isolate subset. Interestingly, this was not the case for the comparator drugs tested, where the non-CF isolate subset was in general more susceptible to each comparator than was the CF isolate subset. Non-CF isolates were 91%, 73.6%, 66.7%, 70.7%, and 84.1% susceptible to amikacin, ceftazidime, levofloxacin (LVX), piperacillin-tazobactam, and tobramycin, respectively, while CF isolates were only 61.2%, 51%, 40.8%, 49%, and 73.5% susceptible to the same antimicrobials, respectively. These results highlight the tendency for CF isolates to be highly antibiotic resistant. Alternatively, meropenem (MEM) susceptibility increased for CF isolates (from 70.2% to 83.7%), while colistin susceptibility remained at 100% for both subsets.

In vitro bactericidal activities of ACHN-975 and LPXC-516.

The in vitro pharmacodynamics of our initial and final lead molecules (ACHN-975 and LPXC-516) were characterized via time-kill assays. Both molecules were assessed against the same 6 clinical isolates of P. aeruginosa, which included two CF isolates and four non-CF strains. The isolates originated from the United States (n = 3), Brazil (n = 1), Vietnam (n = 1), and Taiwan (n = 1) and were selected to provide a range of MIC values to the LpxC inhibitors. When the isolates were susceptible, we used levofloxacin as an active comparator against susceptible isolates due to the concentration-dependent, rapid bactericidal nature of the molecule and because, as in the case of LpxC inhibitors, increases in fluoroquinolone MIC values can arise through single point mutations. Meropenem was used as a control against levofloxacin-resistant strains due to its clinical use for infections caused by MDR P. aeruginosa.

A 3-log₁₀ kill reduction in CFU per milliliter was observed within ≤4 h for all concentrations of both LpxC inhibitors tested (Fig. 2; see also Table S1 in the supplemental material). Rate of bacterial kill increased with increasing concentration, demonstrating that ACHN-975 and LPXC-516 had concentration-dependent bactericidal activity against P. aeruginosa. However, in several cases, regrowth at the 24-h time point was observed, which was correlated with significant increases in MIC values (8- to >32-fold increases in MIC values for some isolates). Upon sequencing, these isolates were found to have various mutations in nxfB, which encodes a repressor of the mexCD-oprJ efflux pump, and mutations in this gene were consistent with those found in the selection experiments described below (17, 18). These phenotypes were stable even after 5 passages on drug-free media, suggesting that there was little to no apparent fitness cost of these mutations. Levofloxacin also demonstrated rapid bactericidal activity against most isolates tested, while meropenem demonstrated more gradual killing (3-log₁₀ kill reduction in CFU per milliliter within ≤24 h) or reduced killing (inability to reduce CFU per milliliter by 3 log₁₀) at the concentration tested and/or within the time frame measured. Representative graphs for 3 isolates are shown in Fig. 2, and the complete results are shown in Table S1.

FIG 2.

Killing kinetics of ACHN-975 and LPXC-516 against six P. aeruginosa isolates. (A) ACHN-975 or levofloxacin vs. P. aeruginosa APAE1136; (B) LPXC-516 or levofloxacin vs. P. aeruginosa APAE1136; (C) ACHN-975 or meropenem vs. P. aeruginosa APAE1232; (D) LPXC-516 or levofloxacin vs. P. aeruginosa APAE1232; (E) ACHN-975 or levofloxacin vs. P. aeruginosa APAE1064; (F) LPXC-516 or levofloxacin vs. P. aeruginosa APAE1064. The APAE1136 MIC values were 0.12 (ACHN-975), 2 (LPXC-516), and 0.5 (levofloxacin) μg/ml; the APAE1232 MIC values were 0.06 (ACHN-975), 0.5 (LPXC-516), and 0.25 (meropenem) μg/ml; and the APAE1064 MIC values were 0.06 (ACHN-975), 0.25 (LPXC-516), and 0.5 (meropenem) μg/ml. Triangles, 4× MIC LpxC inhibitor; upside-down triangles, 8× MIC LpxC inhibitor; solid squares, 16× MIC LpxC inhibitor; asterisks, 32× MIC LpxC inhibitor; open boxes, control 16× MIC.

In vivo bactericidal activity of ACHN-975.

Early in vivo investigations of our LpxC inhibitors suggested that larger, less frequent doses led to maximal efficacy against P. aeruginosa, whereas lower, more frequent doses led to maximal efficacy for Enterobacteriaceae (unpublished data). To confirm this phenomenon and to select doses for future P. aeruginosa in vivo studies, we tested single increasing doses of ACHN-975 in the mouse neutropenic thigh infection model. Colony counts were measured following single doses of ACHN-975 at 5 mg/kg, 10 mg/kg, and 30 mg/kg of body weight against the P. aeruginosa ATCC 27853 strain. A steady reduction in bacterial titers was observed in the first 4 h following treatment for all dosing groups (Fig. 3, top graph), with increasing efficacy associated with increasing total dose. These observations confirmed that future investigations into the in vivo efficacy of LpxC inhibitors against P. aeruginosa should emphasize concentration rather than time-dependent approaches to dosing. For the 5-mg/kg dose, the maximum reduction in bacterial titer was observed at 4 h, with regrowth of bacteria resulting in titers near those observed in vehicle controls at 24 h. However, for both the 10-mg/kg and 30-mg/kg treatment groups, 3-log₁₀ kill was observed by 4 h posttreatment and remained at the limit of detection through 24 h despite no additional compound being administered. PK sampling showed that the level of free drug in this model dropped below the ACHN-975 MIC for this isolate (0.25 μg/ml) by 2 h after treatment with the 10-mg/kg dose and by 4 h after treatment with the 30-mg/kg dose (data not shown). These data confirm the concentration-dependent nature of the pharmacodynamics of this class of LpxC inhibitors.

FIG 3.

(Top) In vivo efficacy of ACHN-975 in a neutropenic mouse thigh model with P. aeruginosa ATCC 27853. Circles, vehicle control; inverted triangles, 5 mg/kg of ACHN-975; triangles, 10 mg/kg of ACHN-975; square, 30 mg/kg of ACHN-975. (Bottom) In vivo efficacy of LPXC-516 in a neutropenic mouse lung model. Closed circles, P. aeruginosa ACH-02 (LPX-516 MIC, 0.5 μg/ml; cystic fibrosis); open circles, P. aeruginosa LES341 (LPX-516 MIC, 0.12 μg/ml; cystic fibrosis); closed squares, P. aeruginosa ACH-04 (LPX-516 MIC, 1 μg/ml); open squares, P. aeruginosa ACH-06 (LPX-516 MIC, 2 μg/ml); inverted triangles, P. aeruginosa ATCC 27853 (LPX-516 MIC, 2 μg/ml).

In vivo efficacy in the mouse neutropenic lung infection model.

The development path of the Achaogen LpxC inhibitor program was focused on the treatment of serious bacterial-associated respiratory infections. Therefore, assessment of efficacy for our LpxC inhibitors was conducted in the neutropenic mouse lung model of P. aeruginosa infection. LPXC-516 was the most extensively evaluated in this model, as it was our reoptimized lead molecule following the clinical failure of ACHN-975. Efficacy was examined against 5 different isolates in this model, including 2 MDR isolates originating from CF patients (Fig. 3, bottom graph). Using a single subcutaneous dose, LPXC-516 was able to achieve stasis and 1-log₁₀ kill against all isolates and 2-log₁₀ kill against 4 of 5 isolates tested. Interestingly, LPXC-516 was more efficacious against isolates derived from patients with CF than from non-CF sources. This phenomenon was consistent across all LpxC inhibitors tested, was not driven by differences in MIC, and remains unexplained.

Assessment of the potential for spontaneous resistance development.

The propensity of bacterial strains to develop resistance to LpxC inhibitors was evaluated in both spontaneous-mutation-frequency and serial-passage experiments. As a new class, LpxC inhibitors have no known resistance mechanism present on mobile elements; thus, these two methods were used to assess the potential for resistance to arise via chromosomal mutation.

Initial investigations focused on the frequency of spontaneous resistance to ACHN-975. We measured the spontaneous-mutation frequency against ACHN-975 in nine strains of P. aeruginosa, including two broadly susceptible lab strains and seven clinical isolates. Mutation frequencies ranged from 1.2 × 10⁻⁷ to 1.5 × 10⁻¹⁰ when selected at 4-fold the MIC of the respective isolates. When the selection conditions were increased to 8-fold the MIC, the mutation frequencies ranged from 1.7 × 10⁻⁸ to undetectable (<1.8 × 10⁻¹¹). The frequency ranges overlapped those observed for levofloxacin, which was tested in parallel for most strains.

A second set of experiments evaluated the spontaneous-mutation frequency of five P. aeruginosa clinical isolates selected from the Micromyx strain collection using an agar plate selection method with LPXC-516, LPXC-289, LPXC-313, or levofloxacin at concentrations 2-, 4-, 8-, and 16-fold the respective strain MICs. The results obtained with ACHN-975 suggested that a relatively high frequency of mutation-based resistance was likely with these molecules, so we therefore included a levofloxacin control group. LPXC-516 had the lowest spontaneous-mutation frequencies, ranging from 5 × 10⁻⁸ to <1 × 10⁻¹⁰ when tested at 4-fold the MIC of the respective isolates and from 7 × 10⁻⁹ to <1 × 10⁻¹⁰ when tested at 8-fold the MIC. LPXC-289, LPXC-313, and levofloxacin had mutation frequencies very similar to each other. At 4-fold the parent strain MICs, these compounds all had at least one isolate with a frequency above the limit of quantification (∼>8 × 10⁻⁸) and at least one isolate with a frequency below the limit of detection (∼<2 × 10⁻¹⁰). At 8-fold the parent strain MICs, LPXC-289, LPXC-313, and levofloxacin had mutation frequencies ranging from highs of 2 × 10⁻⁸, 3 × 10⁻⁸, and 5 × 10⁻⁸, respectively, to a low of <1 × 10⁻¹⁰.

Selection of resistant mutants by passaging.

The potential for resistance development by serial passage was first assessed for ACHN-975 with four strains of P. aeruginosa from the Achaogen culture collection. Cultures were passaged in subinhibitory concentrations of ACHN-975 for 21 passages or until the MIC reached at least 64 μg/ml for three consecutive passages, whichever came first. When the MIC increased at least 4-fold that for the previous passage, the culture was archived for analysis.

Resistance to ACHN-975 increased gradually over time for all strains, but differences between species were observed. The four strains of P. aeruginosa tested were all passaged 21 times without reaching the MIC-stopping criteria. Initial MICs for these strains ranged from 0.12 to 0.5 μg/ml, and the MICs at the end of the experiment ranged from 4 to 32 μg/ml (32- to 128-fold the starting MICs). Levofloxacin MICs increased faster than ACHN-975 MICs for two of the P. aeruginosa strains, reaching the stopping criteria after 10 and 16 passages, with final MICs of 128 and 64 μg/ml, respectively (128-fold above their starting MICs of 1 and 0.5 μg/ml). Levofloxacin MICs against the other two P. aeruginosa strains required 21 passages to increase from 0.5 μg/ml to 64 μg/ml (128-fold).

Serial-passage experiments with LPXC-516, LPXC-289, and LPXC-313 were performed with five isolates of P. aeruginosa from the Micromyx culture collection that were selected to cover a range of MICs (0.5 to 4 μg/ml for LPXC-516, 0.25 to 1 μg/ml for LPXC-289, and 0.25 to 2 μg/ml for LPXC-313). LpxC inhibitor MICs rose at a rate similar to that of levofloxacin MICs when averaged across all strains and inhibitors, but differences could be observed for particular strains and compounds. Specifically, LPXC-516 MICs tended to increase less over 10 passages (1- to 32-fold across the five strains) compared to LPXC-289 (4- to 64-fold increases), LPXC-313 (2- to 128-fold increases), and levofloxacin (8- to 32-fold increases). However, for one strain this pattern was reversed, with the LPXC-516 MIC increasing the most of any compound (16-fold) and the LPXC-313 MIC increasing the least (4-fold). A strain with low starting MICs for all compounds (0.25 to 0.5 μg/ml for the LpxC inhibitors and 0.12 μg/ml for levofloxacin) showed the largest fold increases in MIC (32-fold for LPXC-516 and levofloxacin, 64-fold for LPXC-289, and 128-fold for LPXC-313), while a strain with high starting MICs (1 to 4 μg/ml for the LpxC inhibitors and 1 μg/ml for levofloxacin) showed only 1- to 4-fold MIC increases for the LpxC inhibitors and a 16-fold increase for levofloxacin.

Identification of resistance mechanisms.

Efflux systems were identified as a major contributor to resistance to all four LpxC inhibitors, in agreement with published observations for other LpxC inhibitors (12, 19). Many of the P. aeruginosa spontaneous mutants and regrowth colonies had nonsense, frameshift, or other mutations in nfxB (Table 2), which encodes a repressor of the mexCD-oprJ efflux pump (17, 18). MICs increased 4- to 32-fold in these mutants. These mutants displayed several altered characteristics compared to their respective parents, including slower growth, elevated fluoroquinolone MICs, and collateral sensitivity to imipenem, aztreonam, and amikacin. Less frequently, mutations were found in MexT, a global transcriptional regulator that influences efflux pump expression (20). MexT mutants also showed cross-resistance to fluoroquinolones.

TABLE 2.

Mutations identified in strains resistant to LpxC inhibitors

Locus	Mutation(s)a	Fold MIC shift (compared to parent)
		ACHN-975	LPXC-516	LPXC-289	LPXC-313
nfxB	R21P, R23*, K27insELAE, FS_86, T39P, Q52*, Y101*, E120*, Q130*, T140del14, FS_179, Cys*	16 to ≥64	16	8 to 16	4 to 8
mexS	P59L, L71P, H92Y, FS_94, FS_136, G231S, G244D, Q316*	4 to ≥16	4 to 8	4 to 32	8 to 16
mexT	P242L	8	8	16	16
mexT-mexS intergenic	C insertion −46 bp upstream of mexT	≥16	8	16	16
fabG	V10G, R135S, G150R, A159V, T165I, A167V, L168Q, A169P, N180S, N241R	4	1 to 2	4 to 8	4 to 8
acpP	D57G, D57H	4 to ≥16	2	8 to 16	8 to 16
lpxC upstream	C to G, A, or deleted −11 bp upstream of lpxC	16 to 32	16 to 32	16 to 32	16 to 32

^aAsterisks indicate stop codons. FS, frameshift after indicated amino acid position; ins, insertion following the indicated amino acid; del, deletion following the indicated amino acid.

In line with previous reports, we found that mutation of the cytosine 11 bp upstream of the lpxC start site resulted in elevated MICs (12, 19). We observed new permutations at this site, with the cytosine changed to adenine or guanine or deleted. These mutations were rare, with a frequency of approximately 10⁻¹⁰. In contrast to efflux mutations, these alterations affected all four compounds equally, with 32-fold MIC shifts, regardless of the strain background or mutation type. The mutants did not display cross-resistance to any of the comparators tested (amikacin, aztreonam, ceftazidime, imipenem, levofloxacin, piperacillin-tazobactam, and polymyxin B). This mutation has been associated with increased protein levels of LpxC (12, 19). We found mutations at this locus in strains that lacked preexisting efflux upregulation, including the broadly susceptible strain ATCC 27853, contradicting an earlier hypothesis that efflux upregulation was a necessary precondition for mutation at this site. We did not identify any mutations in the coding sequence of P. aeruginosa lpxC in any of our experiments.

We also found mutations in fabG (Table 2), as has been reported for mutants resistant to the LpxC inhibitor CHIR-090 (19). The FabG enzyme is involved in fatty acid biosynthesis, a pathway that shares intermediates with the LPS biosynthesis pathway. Furthermore, we identified mutations in acpP, which encodes acyl carrier protein, an essential protein that participates in multiple steps of lipid synthesis, including direct interaction with FabG (21). Our fabG and acpP mutants all produced small colonies on inhibitor-free agar plates, consistent with the growth defect observed for earlier fabG mutants, but they also produced colonies of normal size on plates containing LpxC inhibitor at 4-fold above the MIC of the parent strain (19). Mutants with these various resistance mechanisms were found in colonies selected from both passage selection and spontaneous-mutation-frequency studies.

DISCUSSION

The LpxC protein is an intriguing target for a new antibacterial class because it is an essential enzyme in P. aeruginosa and other Gram-negative bacteria and because its role in lipid A biosynthesis offers a unique mechanism of action that does not overlap currently available antibiotics. Because LpxC is a new target, there are no known resistance genes on mobile elements, and preexisting resistance on the bacterial chromosome is expected to be minimal. Previous in vitro and in vivo studies on the activity of small-molecule LpxC inhibitors against important pathogens demonstrated promise for this target (22).

ACHN-975, LPXC-516, LPXC-313, and LPXC-289 are potent inhibitors of P. aeruginosa, including MDR strains as well as those isolated from CF patients. We found that these LpxC inhibitors are rapidly bactericidal in vitro and that this activity translated to in vivo efficacy against P. aeruginosa. We found that ACHN-975, LPXC-516, LPXC-313, and LPXC-289 shared some resistance liabilities with LpxC inhibitors described by others. We demonstrated that a mutation in the cytosine 11 bp upstream of lpxC resulted in a 32-fold MIC increase. We hypothesize, based on our results and those of previous investigators, that this mutation would affect any LpxC inhibitor equally and thus supersede any impact of efflux upregulation. However, it is important to note that these mutations were rare in our studies (frequency, 1 in 10¹⁰).

The rate of resistance development and the magnitude of the MIC changes described for the molecules from the Achaogen program made finding a dose that had the potential to prevent emergence of clinical resistance critical to moving the program back into clinical development. However, the compounds described in this paper could not be further developed due to an inability to find such a dosing regimen that achieved efficacious exposures while reducing the potential for resistance development and minimizing toxicity signals. However, improvements in the toxicity profile of these drugs might be possible such that higher exposures could be achieved that are safe while preventing outgrowth of resistant mutants. Overall, the data presented here support the continued pursuit of LpxC as a target for a new class of antibacterials against concerning Gram-negative pathogens.

MATERIALS AND METHODS

Bacterial strains.

P. aeruginosa (n = 250) clinical respiratory isolates, including 49 isolates from CF patients, were collected between 2010 and 2016 and were assessed at International Health Management Associates (IHMA; Shaumsburg, IL) in MIC assays. A total of 39.6% of these isolates originated from North America, including all CF isolates; 20.4% were from Europe, and the remainder, 40%, came from locations throughout the rest of the world. P. aeruginosa isolates investigated in the time-kill study came from the following sources: (i) APAE1148 (USA; 2005) and APAE1136 (Brazil; 2005) were provided by JMI Labs, North Liberty, IA; (ii) APAE1019 and APAE1064 were from CF patients, provided by Children’s Hospital, Seattle, WA; and (iii) APAE1260 (Vietnam; 2012) and APAE1262 (Taiwan; 2012) were provided by IHMA, Schaumsburg, IL. APAE001 is P. aeruginosa ATCC 27853.

Antimicrobial compounds.

LpxC inhibitors ACHN-975, LPXC-516, LPXC-313, and LPXC-289 were synthesized at Achaogen, Inc., as described by Cohen et al. (13). LpxC inhibitor stock solutions were prepared in 100% dimethyl sulfoxide (DMSO) and stored at –20°C. For time-kill experiments, levofloxacin (Waterstone Technologies) and meropenem (TCI Chemicals) stock solutions were prepared in water and stored at room temperature and –80°C, respectively, until use. For MIC experiments conducted at IHMA, the source of the comparator antibiotics was as follows: USP for amikacin, ceftazidime, colistin, meropenem, tazobactam, and tobramycin, Merck for ertapenem, and Sigma for levofloxacin and piperacillin.

MIC assays.

MICs were determined by broth microdilution in cation-adjusted Mueller-Hinton broth (CAMHB) according to Clinical and Laboratory Standards Institute (CLSI) guidelines (23). For LpxC inhibitor MICs, all assay wells, including the inhibitor-free growth control, contained ≤1% DMSO. CLSI criteria were used to interpret the percent susceptibility for all comparator antimicrobials with the exception of colistin, for which EUCAST criteria were utilized.

LpxC enzyme inhibition assay.

The LpxC enzyme inhibition assay was adapted from the method detailed by Hale et al. (24), which uses liquid chromatography-mass spectroscopy (LC-MS) to measure the fraction of substrate converted to product. Recombinant P. aeruginosa LpxC purified from E. coli was utilized for the assay. Test LpxC inhibitors previously dissolved in DMSO were diluted to a 10× stock concentration in water such that the final DMSO concentration in the assay was less than 1%. Three microliters of 10× test compound was then aliquoted in 3-fold dilutions across the row of a 96-well plate. Twenty-four microliters of a buffered LpxC enzyme-containing solution (fresh enzyme was obtained from a −80°C previously aliquoted stock) was then added to the test compound dilutions. The enzyme and test compound were allowed to incubate for 10 min at room temperature. To initiate the reaction, 3 μl of 400 μM UDP-3-O-(R-3-hydroxydecanoyl)-N-acetylglucosamine (CAS number 953426-26-1; obtained from Beijing Honghui Meditech Co., Ltd.) was added. The final reaction conditions of the assay were 10 mM sodium phosphate (pH 7.5), 0.005% Triton X-100, 1 mM Tris(2-carboxyethyl)phosphine hydrochloride (TCEP), 40 μM UDP-3-O-(R-3-hydroxydecanoyl)-N-acetylglucosamine, and 10 μM Zn(acetate)₂. The final expected concentration of catalytically active LpxC was estimated to be approximately 0.1 nM. The reaction was allowed to progress to approximately 20% completion in the absence of test compound over 8 h, at which point 3 μl of 1 M HCl was added to the reaction to terminate enzyme activity.

In vitro time-kill study.

ACHN-975 and LPXC-516 were assessed against strains APAE1136 (non-CF; versus LVX), APAE1148 (non-CF; versus MEM), APAE1232 (non-CF; versus MEM), APAE1260 (non-CF; versus LVX), APAE1064 (CF; versus MEM), and APAE1019 (CF; versus MEM) in in vitro time-kill assays. Each strain was grown to log phase via agitation at 37°C in CAMHB, and then the bacterial culture was split into 10-ml volumes. The culture was then mixed with the antibiotic at the concentration of interest and returned to incubation at 37°C with agitation. ACHN-975 and LPXC-516 were tested at 4-, 8-, 16-, and 32-fold concentrations above the individual strain MIC, while each comparator was tested at a concentration 16-fold the strain MIC. The MIC of each antimicrobial was determined in MIC assays, as previously described. The starting inoculum was determined from an untreated growth control at the 0-h time point, the time at which the culture was split. Aliquots were removed from each culture 1, 2, 4, 6, and 24 h of incubation.

Viability of the cultures over time was monitored by serially diluting samples in sterile saline and spreading 0.1 ml of the diluted samples onto agar plates. For some experiments, the lowest dilution plated was 10⁻¹, while for others, undiluted cultures were also plated, resulting in different limits of detection. Agar plates were incubated overnight at 35°C in ambient air. The following day, CFU were counted from plates with well-separated colonies using an aCOLyte Synoptics colony counter and converted to CFU per milliliter using the appropriate dilution factor. Bactericidal activity was defined as a ≥3-log₁₀ decrease in the number of viable cells in a culture, equivalent to 99.9% killing of the inoculum (25).

In vivo time-kill assay.

Neutropenic ICR mice were utilized for this study. Groups of 5 mice were rendered neutropenic with intraperitoneal administration of 150 mg/kg and 100 mg/kg of cyclophosphamide at 4 days and 1 day prior to infection, respectively. The right thighs of individual animals were inoculated with 10 (24) CFU P. aeruginosa strain ATCC 27853 via intramuscular (i.m.) injection. A single dose of 5, 10, or 30 mg/kg of ACHN-975 was administered 2 h postinfection. At 0, 2, 4, 8, 16, and 24 h after treatment, animals were humanely euthanized with CO₂ asphyxiation, and the muscle of the right thigh was harvested from each test animal and homogenized in 5 ml of phosphate-buffered saline (PBS; pH 7.4) with a Polytron homogenizer.

All aspects of this work, including housing, experimentation, and animal disposal, were performed in general accordance with the Guide for the Care and Use of Laboratory Animals (26). The study protocol was approved by the Eurofins Panlabs IACUC. The experiment was performed under animal biosafety level 2 (ABSL2) conditions in the Eurofins Pharmacology Discovery Services AAALAC-accredited vivarium with the oversight of veterinarians to ensure compliance with the institution’s IACUC regulations and the humane treatment of laboratory animals.

Mouse neutropenic lung infection model.

Neutropenic ICR mice were utilized for this study. Groups of 5 mice were rendered neutropenic with intraperitoneal administration of 150 mg/kg and 100 mg/kg of cyclophosphamide at 4 days and 1 day prior to infection, respectively. On day 0, animals were inoculated intranasally (0.02 ml/lung) with the pertinent organism after anesthesia with propofol (∼30 mg/kg intravenously [i.v.]) or etomidate (Lipuro) emulsion (20 mg/kg i.v.). A single 0.45-, 1.5-, 4.5-, 15-, 45-, 150-, or 450-mg/kg dose of C051516 or a vehicle was administered 2 h postinfection. At 26 h after inoculation, animals were euthanized with CO₂ asphyxiation and the lung tissue was harvested from each of the test animals. The tissues were homogenized in 1 ml of PBS (pH 7.4) with a homogenizer. Homogenates (0.1 ml) were used for serial 10-fold dilutions and plated onto MacConkey II agar for colony counts. Homogenates (0.1 ml) were diluted serially 10-fold and plated onto Trypticase soy broth in 1.5% Bacto agar for colony count (CFU per gram) determination. Plates were incubated at 35°C overnight prior to colony count.

All aspects of this work, including housing, experimentation, and animal disposal, were performed in general accordance with the Guide for the Care and Use of Laboratory Animals (26). The study protocol was approved by the Eurofins Panlabs IACUC. The experiment was performed under ABSL2 conditions in the Eurofins Pharmacology Discovery Services AAALAC-accredited vivarium with the oversight of veterinarians to ensure compliance with the institution’s IACUC regulations and the humane treatment of laboratory animals.

Spontaneous-mutation-frequency studies.

For studies with ACHN-975, selection plates were prepared by adding antibiotics to molten agar at ≤50°C. For each strain, 5 separate colonies were picked from fresh agar streaks and used to inoculate 5 independent cultures in CAMHB. The cultures were grown overnight and then diluted into fresh media. When cultures reached a density of approximately 10⁹ CFU/ml, they were pelleted by centrifugation and resuspended in 1/10 the original volume, for a density of approximately 10¹⁰ CFU/ml. Samples of the cell suspensions (0.2 ml) were spread onto agar plates containing either 4× or 8× MIC of inhibitor. The inoculum density was determined by serially diluting the cell suspension and plating to inhibitor-free agar. Plates were incubated at 35°C in ambient air. CFU on the inoculum control plates were counted after 24 h. CFU on antibiotic plates were inspected at 24 h, and CFU were counted after 48 h. Colony counts from the 5 independent cultures for each antibiotic concentration were compared, and if any were >10-fold higher than the average of the other plates, they were excluded from the frequency calculation.

Resistant colonies were archived as frozen glycerol stocks. Stocks were streaked onto antibiotic-free agar plates for 5 passages to test for stability of the resistance phenotype. Initial and passaged strains were analyzed alongside the parental strains by MIC assay to determine the magnitude of resistance.

For studies with LPXC-516, LPXC-289, and LPXC-313, dense bacterial suspensions were prepared in CAMHB by resuspending colonies selected from freshly streaked agar plates. Inoculum densities were adjusted so that 10⁹ to 10¹⁰ CFU were applied to plates when streaking 0.3 ml of the suspension. CAMH agar plates containing 2×, 4×, 8×, 16×, or 32× MIC of inhibitor (ACHN-975, LPXC-516, LPXC-289, and LPXC-313) were used. The inoculum density was determined on inhibitor-free agar. CFU on the inoculum control plates were determined after 24 h of incubation, while the CFU on antibiotic plates were counted at 48 h.

Serial passage at subinhibitory concentrations.

In the case of studies with ACHN-975, for serial-passage studies, bacteria were exposed to a 2-fold dilution series of antibiotic in CAMHB, and cells that survived exposure to one-half the MIC were used as the inoculum for the subsequent culture day. The assay was performed in 1-ml volumes in plastic culture tubes. Each day, the MIC was determined as the concentration that prevented visible growth. The culture containing half the MIC was diluted 4-fold into sterile saline and vortexed briefly, and then 20 μl of the cell suspension was used to inoculate a fresh antibiotic dilution series. The inoculum culture was streaked onto Mueller-Hinton agar to verify culture purity. Whenever an MIC increase occurred, colonies from the agar plate were stored as frozen glycerol stocks for future analysis. Passaging continued until (i) the MIC was at or above a prespecified elevated concentration for 3 continuous passages or (ii) for 21 total passages, whichever was sooner.

In the case of studies with LPXC-516, LPXC-289, and LPXC-313, a modified broth microdilution MIC methodology was used. Briefly, a standard MIC was conducted per CLSI methods. After incubation at 35°C for 16 to 20 h, plates were read and the MICs determined.

After reading the MIC plates from the prior passage, the entire inoculum (200 μl) was removed from the well representing 0.5-fold the MIC for each drug and was added to separate sterile tubes containing 3 ml of CAMHB. If the well below the MIC was judged to contain only marginal growth, the entire contents (200 μl) were also removed from the well representing 0.25-fold the MIC and combined with the 0.5-fold growth in 3 ml of Mueller-Hinton broth II (MHBII). The tubes were incubated at 35°C with shaking at 150 rpm. Growth in each tube was monitored with a turbidity meter, and after sufficient regrowth, a suspension equivalent to a 0.5 McFarland standard was made. The standardized suspension was then diluted 1:20 in MHBII and used as the inoculum for the subsequent passage for that particular drug. The inoculum was hand pipetted (10 μl/well) into a freshly thawed MIC plate to achieve approximately 5 × 10⁵ CFU/ml in the appropriate wells. Plates were incubated at 35°C for 16 to 20 h and the MIC for the next passage was read as described above. Following each inoculation, a 1-ml aliquot of each inoculum was mixed with cryoprotectant and frozen at –80°C for future analysis.

The entire process was repeated through 10 serial passages and the final MIC values were compared to those of the parent isolates.

Identification of resistance mutations.

Resistant mutants were analyzed by PCR followed by Sanger sequencing to identify mutations conferring resistance in lpxC, fabG, and nfxB, including upstream and downstream regions. Isolates were then subjected to whole-genome sequencing. Parental isolates that lacked a published genome were sequenced using a Pacific Biosciences Sequel system to generate reference genomes. Demultiplexed reads were filtered using barcode quality score (>0.45), and filtered reads were corrected, trimmed, and assembled using Canu. The raw assembly was then polished against the consolidated subreads using PacBio’s Arrow algorithm. Resistant strains were sequenced using Illumina HiSeq, paired end 2 × 150 bp. Reads were mapped to the reference genomes in Geneious v10.0.9 (Biomatters Ltd.).